[RFC PATCHv2 09/20] ARM: architected timers: add A15 architected timers

[Marc Zyngier]

On Fri, 2011-04-08 at 19:47 -0700, Stephen Boyd wrote:

Hi Stephen,

> On 3/15/2011 8:12 AM, Marc Zyngier wrote:
> >
> > diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
> > index b7e504b..38ee56e 100644
> > --- a/arch/arm/Kconfig
> > +++ b/arch/arm/Kconfig
> > @@ -1407,9 +1407,14 @@ config HOTPLUG_CPU
> >  	  Say Y here to experiment with turning CPUs off and on.  CPUs
> >  	  can be controlled through /sys/devices/system/cpu.
> >  
> > +config HAVE_ARCH_TIMERS
> > +	bool
> > +	depends on CPU_V7
> > +	select TICK_ONESHOT
> 
> I don't think you need to select TICK_ONESHOT. The user should be
> selecting NO_HZ or HIGH_RES_TIMERS instead, right?

The A15 timer only has the one-shot feature (ie no auto-relead). Not
having TICK_ONESHOT would render the timer completely unusable if
neither NO_HZ nor HIGH_RES_TIMERS are selected. For that reason, I
believe selecting TICK_ONESHOT is necessary.

> > +
> > +static struct clocksource clocksource_counter = {
> > +	.name	= "arch_sys_counter",
> > +	.rating	= 400,
> > +	.read	= arch_counter_read,
> > +	.mask	= CLOCKSOURCE_MASK(56),
> > +	.flags	= (CLOCK_SOURCE_IS_CONTINUOUS | CLOCK_SOURCE_VALID_FOR_HRES),
> 
> I'm fairly certain that CLOCK_SOURCE_VALID_FOR_HRES is a private flag
> not to be used by clocksource drivers. At least grepping the kernel for
> it shows no results outside of kernel/time/

You're probably right here. Will fix.

Thanks for reviewing.

	M.
-- 
Reality is an implementation detail.